← All work

OBD2Tech Capture + Journal

A read-only journal for one Jeep that never guesses past its evidence.

A read-only diagnostic system for my own Jeep that compares the vehicle against its own history, traces each finding to its source recording or scan, and answers 'Unknown' when the evidence cannot support a conclusion.

The idea

OBD2Tech starts with one owner and one 2014 Jeep Cherokee showing a catalyst code. A code reader can name P0430 in seconds, but it cannot say whether the catalyst is failing or the oxygen sensor watching it is part of the story, and two drives are easily mistaken for a trend. The frustration is confident answers built on thin evidence, with no way to check where a conclusion came from.

When it works, the Jeep is measured against its own past rather than a generic spec sheet. A phone records drives and parked scans without changing anything on the car, each bundle is signed and checked again on the Mac, and the Journal says what changed and how strong the evidence is. The insight is that restraint is the feature: when the evidence cannot support a conclusion, the answer is Unknown.

Further out, Drive Coach is meant to offer driving advice only after it passes replay tests on real drives. Read-only testing on a second vehicle, a Ford, is due to start soon, one reviewed vehicle profile at a time; a platform that fully supports other vehicles is not built yet. Today the Journal holds 149 accepted sessions and 546 analysis results, and the iPhone recorder is installed but still has real-vehicle checks to pass before it replaces the Android backup.

How it works

  1. Vehicle signal
  2. iPhone capture
  3. Signed bundle
  4. Journal

The iPhone app reads a Bluetooth OBD adapter through a read-only, exclusive lease and seals each run into a signed bundle. Bundles travel through CloudKit to a Mac relay and ingest app. The Journal, a Python engine with a read-only local dashboard, records provenance, runs immutable analyses, and maps findings onto a 3D vehicle and transmission model.

Where it stands today

The Journal holds 149 accepted sessions and 546 immutable analysis results, and it adopted Vehicle Evidence on 22 September 2026. The iPhone recorder, version 1.1.13, is installed and signs its field evidence but is not yet the accepted primary recorder; the Android recorder passed real-vehicle acceptance and remains the backup. With P0430 present, OBD2Tech identifies Bank 2 upstream oxygen-sensor behavior as a confounder worth testing before a catalyst-only conclusion, while also finding that it does not fully explain the measured bank disparity.

Working now

  • iPhone drive capture and parked deep scans
  • Signed evidence bundles with Mac receipts
  • Journal briefing, diagnostics, and drive-coach research
  • 3D garage that maps findings to vehicle zones
  • Android backup recorder

Not there yet

  • The iPhone recorder is not yet the accepted primary recorder: real-vehicle gates remain open, including a warm daily and Drive Coach V2 run, offline deep-scan parity, an extended four-hour capture, a new cellular upload, and helper restart recovery.
  • The 3D garage is a source-calibrated blockout, not an exact CAD model, and the system is built around one owner's vehicle.

How it got here

  1. Started as a Python OBD-II toolkit and vehicle emulator for a P0430 catalyst trouble code on a 2014 Jeep Cherokee
  2. An Android phone became the field recorder
  3. The Android recorder passed 36 of 36 real-vehicle routes
  4. Native iPhone recorder built
  5. Vehicle Evidence and signed field plans adopted

What comes next

Next step

Record one real Drive Coach V2 acceptance drive.

Later ideas, not current features

  1. Accept the iPhone recorder as primary after the remaining real-vehicle cutover checks
  2. Qualify the odometer reading (an addition, not a cutover gate)
  3. Begin read-only testing on a second vehicle, a Ford
  4. Produce a public-safe parser and simulator subset

What I checked, and what stays bounded

Dated evidence

  1. Android real-vehicle acceptance: 36 of 36 routesphysical
  2. Attended hardware capture and device workflowphysical
  3. Vehicle Evidence adopted: 546 analysis results across 149 sessionsruntime
  4. iPhone recorder 1.1.13 recovered all 57 signed field plansrelease

Boundaries it keeps

  • The VIN stays private, and no GPS or route data is collected
  • Diagnostic support is not a repair guarantee
  • Capture and journal authorities remain distinct
  • The Journal dashboard is read-only

My role

Designed the hardware-facing capture flow, transport controls, evidence receipts, journal model, and recovery procedures; directed the implementation and validated it against the owner's vehicle and simulator paths.